Fewer false flags, more organized-scheme catches
Graph-based link analysis with explainable risk tiers reducing false referrals 31% and lifting hit rate 22%.
Rules catch individual red flags; organized fraud rings exploit the gaps between them.
The Challenge
A national TPA processed 400,000+ claims annually with a rules-based fraud flagging system that produced a 68% false positive rate. SIU investigators spent the majority of their time dismissing false referrals rather than pursuing actionable cases. Meanwhile, organized fraud schemes involving coordinated claimants, staged accidents, and complicit vendors were identified only after significant payouts—often by external audit rather than internal detection.
The Innovoco Solution
We implemented graph-style link analysis across parties, vendors, and historical claims with explainable risk tiers and playbook-driven referrals. The system surfaces relationship patterns that rules-based approaches miss while providing investigators with visual evidence and structured case summaries.
Phase 1 — Graph construction and historical analysis
Built a claims knowledge graph linking claimants, adjusters, attorneys, medical providers, and repair vendors across 5 years of claims history. Identified known fraud patterns (rings, staged losses, provider mills) and calibrated anomaly detection against confirmed SIU outcomes.
Phase 2 — Real-time scoring and investigator workflow
Deployed real-time link analysis on new claims as they enter the system. Risk tiers with explainable features route referrals to SIU with visual relationship maps and structured case summaries. Investigators provide outcome feedback that improves model precision quarterly.
Key implementations
Claims knowledge graph
Entities (claimants, providers, vendors, vehicles, addresses, phone numbers) linked across millions of claims with relationship strength and recency scoring.
Explainable risk tiers
Each referral includes the specific relationship patterns, anomaly signals, and historical context that drove the risk score—not a black-box number.
Visual investigation maps
Interactive graph visualizations let investigators explore connections, filter by relationship type, and drill into individual claims within a network.
Playbook-driven referrals
Different fraud typologies (staged loss, provider mill, identity fraud) generate tailored investigation checklists with recommended evidence collection steps.
Outcome feedback loop
Investigator disposition (confirmed fraud, suspicious, cleared) feeds back to model training, improving precision with each quarterly update.
Technical Innovation
Relationship mapping detects clusters of connected entities that traditional flat-file analysis cannot see. Every network-level risk score decomposes into human-readable relationship evidence—critical for SIU teams that must document rationale for every investigation opened.
Impact
- 31% reduction in false referrals, freeing investigator capacity for actionable cases.
- 22% improvement in fraud hit rate on referred cases.
- Model refresh cycle under 8 weeks, incorporating latest confirmed outcomes.
- Organized-scheme detection identified three multi-million-dollar rings in the first year that rules-based systems had missed.
Investigators spend their time on cases that matter—with visual evidence and structured summaries that accelerate investigation and support prosecution or recovery.
Explore this outcome on your stack
We map scope, guardrails, and rollout to your data boundaries and teams—practical next steps, not a generic slide deck.
60 min · Free · No obligation
More case studies
Explore adjacent patterns from the same production playbook.
AI-Powered Business Intelligence
Ask questions in plain language, get governed answers with charts — no analysts, no ticket queue.
Read more
Customer support & voice
Triage, sentiment-aware routing, CRM-grounded answers at scale.
Read more
Regulated onboarding (KYC / AML)
Document intelligence, exception queues, immutable evidence for auditors.
Read more